What is ML – Really?

Learning vs Coding

~15 min

Now you'll see the difference yourself. Below are two systems trying to predict rain - one using rules, one using machine learning.

ℹ️ What you'll do

Test both systems with different weather scenarios. Pay special attention to the edge cases marked with ⚠️ - these are scenarios that break simple rules.

Input Data:
Temperature
25°C
Humidity
85%
Actual Weather
🌧️ Rain

What Did You Notice?

After trying different scenarios, you probably noticed:

  • Rules are predictable - Same input always gives same output
  • Rules break on edge cases - “Cold BUT humid” confuses simple logic
  • ML adapts to patterns - It learned from data rather than explicit rules
  • ML isn't perfect either - It can make mistakes too!

Coding an Answer

  • • You write explicit if/else logic
  • • Every rule must be thought through
  • • Easy to understand and debug
  • • Breaks on unexpected input
  • • Requires updates for new patterns

Learning an Answer

  • • Computer finds patterns in data
  • • Handles complexity automatically
  • • Harder to interpret why it decided
  • • Adapts to new situations
  • • Quality depends on training data

⚠️ Important Insight

Neither approach is “better” - they're tools for different problems. Use rules when logic is simple and clear. Use ML when patterns are complex or hidden.

💡 When to Use Each Approach

  • Use Rules: Login validation, age verification, simple calculations
  • Use ML: Image recognition, language translation, recommendation systems
  • Use Both: Many real systems combine rules (for safety/compliance) and ML (for complex decisions)